Evidence Hierarchy Framework

Use This Skill When

• The user wants to know whether a thesis is built on evidence or on narrative.
• The analysis contains many claims with very different levels of support.
• The task is to rank evidence quality and expose where confidence is overstated.
• The goal is to decide which parts of the thesis are investable now and which parts still require work.

Required Inputs

• The thesis or set of claims being tested.
• Source materials such as filings, transcripts, KPIs, third-party data, channel checks, expert calls, or survey results.
• Time frame for the evidence: current quarter, recent year, or longer history.
• Any contested claims or areas where the user has low confidence.
• Optional but useful: prior theses, management promises, and outcome data for back-checking.

Workflow

1. Break the thesis into discrete claims. Each claim should be testable and important to valuation or timing.
2. Inventory the evidence. Assign each claim the best available evidence set, including direct company data, third-party confirmation, management statements, and inference.
3. Rank evidence quality. Judge each source by directness, reliability, freshness, independence, and falsifiability.
4. Compare claim importance to evidence strength. Highlight where weak evidence is carrying a large share of the thesis weight.
5. Build the hierarchy. Separate high-conviction claims, provisional claims, and unsupported claims.
6. Recommend next research steps. Say what evidence would most improve or weaken conviction.

Output Requirements

• Evaluate claims individually rather than giving one blanket confidence score.
• Be explicit about the difference between observed fact, informed inference, and narrative assumption.
• Penalize stale evidence and management-only support when there is no external confirmation.
• End with the strongest supported claim, the weakest link, and the highest-value next check.
• Keep the framework useful for buy-side decision-making, not academic source taxonomy.

Output Template

Quality Checks

• Every important thesis claim is mapped to actual evidence.
• Evidence quality is judged on both rigor and recency.
• The framework distinguishes source quality from claim importance.
• Claims with weak support are not allowed to drive a strong conclusion.
• The final answer tells the reader what is truly known versus assumed.

Guardrails

• Do not equate management commentary with independent validation.
• Do not hide weak evidence behind confident prose.
• Avoid binary "proven / unproven" labels when evidence is mixed; state degree of confidence.
• Do not overweight anecdotal channel checks unless they are broad, recent, and triangulated.
• If evidence conflicts, show both sides and explain which source deserves more weight.

Example Prompts

• Rank the evidence behind this investment thesis and tell me which claims are actually supported.
• Separate facts, strong inferences, and weak assumptions in this long thesis.
• Build an evidence hierarchy for these claims and show me the weakest link.